安装llama-cpp-python踩坑记

try2find2025-07-04 21:34

1. 安装llama-cpp-python报错

python 复制代码 
$ CMAKE_ARGS="-DLLAMA_CUBLAS=on" pip install llama-cpp-python -i https://pypi.tuna.tsinghua.edu.cn/simple
Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
Collecting llama-cpp-python
  Downloading https://pypi.tuna.tsinghua.edu.cn/packages/de/6d/4a20e676bdf7d9d3523be3a081bf327af958f9bdfe2a564f5cf485faeaec/llama_cpp_python-0.3.9.tar.gz (67.9 MB)
     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 67.9/67.9 MB 5.3 MB/s eta 0:00:00
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
  Installing backend dependencies ... done
  Preparing metadata (pyproject.toml) ... done
Requirement already satisfied: typing-extensions>=4.5.0 in /home/wuwenliang/anaconda3/envs/llmtuner/lib/python3.10/site-packages (from llama-cpp-python) (4.13.2)
Requirement already satisfied: numpy>=1.20.0 in /home/wuwenliang/anaconda3/envs/llmtuner/lib/python3.10/site-packages (from llama-cpp-python) (1.26.4)
Collecting diskcache>=5.6.1 (from llama-cpp-python)
  Downloading https://pypi.tuna.tsinghua.edu.cn/packages/3f/27/4570e78fc0bf5ea0ca45eb1de3818a23787af9b390c0b0a0033a1b8236f9/diskcache-5.6.3-py3-none-any.whl (45 kB)
Requirement already satisfied: jinja2>=2.11.3 in /home/wuwenliang/anaconda3/envs/llmtuner/lib/python3.10/site-packages (from llama-cpp-python) (3.1.6)
Requirement already satisfied: MarkupSafe>=2.0 in /home/wuwenliang/anaconda3/envs/llmtuner/lib/python3.10/site-packages (from jinja2>=2.11.3->llama-cpp-python) (3.0.2)
Building wheels for collected packages: llama-cpp-python
  Building wheel for llama-cpp-python (pyproject.toml) ... error
  error: subprocess-exited-with-error
  
  × Building wheel for llama-cpp-python (pyproject.toml) did not run successfully.
  │ exit code: 1
  ╰─> [29 lines of output]
      *** scikit-build-core 0.11.5 using CMake 3.22.1 (wheel)
      *** Configuring CMake...
      loading initial cache file /tmp/tmp01d6kko6/build/CMakeInit.txt
      -- The C compiler identification is GNU 11.2.0
      -- The CXX compiler identification is GNU 11.2.0
      -- Detecting C compiler ABI info
      -- Detecting C compiler ABI info - done
      -- Check for working C compiler: /usr/bin/gcc - skipped
      -- Detecting C compile features
      -- Detecting C compile features - done
      -- Detecting CXX compiler ABI info
      -- Detecting CXX compiler ABI info - done
      -- Check for working CXX compiler: /usr/bin/g++ - skipped
      -- Detecting CXX compile features
      -- Detecting CXX compile features - done
      -- Found Git: /usr/bin/git (found version "2.34.1")
      CMake Error at vendor/llama.cpp/CMakeLists.txt:108 (message):
        LLAMA_CUBLAS is deprecated and will be removed in the future.
      
        Use GGML_CUDA instead
      
      Call Stack (most recent call first):
        vendor/llama.cpp/CMakeLists.txt:113 (llama_option_depr)
      
      
      -- Configuring incomplete, errors occurred!
      See also "/tmp/tmp01d6kko6/build/CMakeFiles/CMakeOutput.log".
      
      *** CMake configuration failed
      [end of output]
  
  note: This error originates from a subprocess, and is likely not a problem with pip.
  ERROR: Failed building wheel for llama-cpp-python
Failed to build llama-cpp-python
ERROR: Failed to build installable wheels for some pyproject.toml based projects (llama-cpp-python)

安装报错，分析如下：

这个错误是因为 LLAMA_CUBLAS 选项已经被弃用，建议使用 GGML_CUDA 替代。你需要修改安装命令中的 CMake 参数。

完整安装命令（包含更多可能的依赖）：

python 复制代码 
CMAKE_ARGS="-DGGML_CUDA=on -DCMAKE_CUDA_COMPILER=/usr/local/cuda/bin/nvcc" pip install llam

安装仍旧失败：

python 复制代码 
  collect2: error: ld returned 1 exit status
  ninja: build stopped: subcommand failed.

  *** CMake build failed
  error: subprocess-exited-with-error
  
  × Building wheel for llama-cpp-python (pyproject.toml) did not run successfully.
  │ exit code: 1
  ╰─> See above for output.
  
  note: This error originates from a subprocess, and is likely not a problem with pip.
  full command: /llmtuner/bin/python /llmtuner/lib/python3.10/site-packages/pip/_vendor/pyproject_hooks/_in_process/_in_process.py build_wheel /tmp/tmp2wtlptdn
  cwd: /tmp/pip-install-jm38bxck/llama-cpp-python_69bdce92ceff4d4db8aec6e07d4f05e3
  Building wheel for llama-cpp-python (pyproject.toml) ... error
  ERROR: Failed building wheel for llama-cpp-python
Failed to build llama-cpp-python
ERROR: Failed to build installable wheels for some pyproject.toml based projects (llama-cpp-python)

2. 采用源码安装

python 复制代码 
git clone --recursive https://github.com/abetlen/llama-cpp-python.git
cd llama-cpp-python
CMAKE_ARGS="-DGGML_CUDA=on" pip install . -i https://pypi.tuna.tsinghua.edu.cn/simple

仍旧失败

python 复制代码 
$ CMAKE_ARGS="-DGGML_CUDA=on" pip install . -i https://pypi.tuna.tsinghua.edu.cn/simple
Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
Processing /home/software/llama-cpp-python
  Installing build dependencies ... done
  Getting requirements to build wheel ... done
  Installing backend dependencies ... done
  Preparing metadata (pyproject.toml) ... done
Requirement already satisfied: typing-extensions>=4.5.0 in /llmtuner/lib/python3.10/site-packages (from llama_cpp_python==0.3.9) (4.13.2)
Requirement already satisfied: numpy>=1.20.0 in /llmtuner/lib/python3.10/site-packages (from llama_cpp_python==0.3.9) (1.26.4)
Collecting diskcache>=5.6.1 (from llama_cpp_python==0.3.9)
  Using cached https://pypi.tuna.tsinghua.edu.cn/packages/3f/27/4570e78fc0bf5ea0ca45eb1de3818a23787af9b390c0b0a0033a1b8236f9/diskcache-5.6.3-py3-none-any.whl (45 kB)
Requirement already satisfied: jinja2>=2.11.3 in /llmtuner/lib/python3.10/site-packages (from llama_cpp_python==0.3.9) (3.1.6)
Requirement already satisfied: MarkupSafe>=2.0 in /llmtuner/lib/python3.10/site-packages (from jinja2>=2.11.3->llama_cpp_python==0.3.9) (3.0.2)
Building wheels for collected packages: llama_cpp_python
  Building wheel for llama_cpp_python (pyproject.toml) ... error
  error: subprocess-exited-with-error
  
  × Building wheel for llama_cpp_python (pyproject.toml) did not run successfully.
  │ exit code: 1
  ╰─> [64 lines of output]
      *** scikit-build-core 0.11.5 using CMake 3.22.1 (wheel)
      *** Configuring CMake...
      loading initial cache file /tmp/tmpwhq61gdi/build/CMakeInit.txt
      -- The C compiler identification is GNU 11.2.0
      -- The CXX compiler identification is GNU 11.2.0
      -- Detecting C compiler ABI info
      -- Detecting C compiler ABI info - done
      -- Check for working C compiler: /usr/bin/gcc - skipped
      -- Detecting C compile features
      -- Detecting C compile features - done
      -- Detecting CXX compiler ABI info
      -- Detecting CXX compiler ABI info - done
      -- Check for working CXX compiler: /usr/bin/g++ - skipped
      -- Detecting CXX compile features
      -- Detecting CXX compile features - done
      -- Found Git: /usr/bin/git (found version "2.34.1")
      -- Looking for pthread.h
      -- Looking for pthread.h - found
      -- Performing Test CMAKE_HAVE_LIBC_PTHREAD
      -- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Success
      -- Found Threads: TRUE
      -- ccache found, compilation results will be cached. Disable with GGML_CCACHE=OFF.
      -- CMAKE_SYSTEM_PROCESSOR: x86_64
      -- Including CPU backend
      -- Found OpenMP_C: -fopenmp (found version "4.5")
      -- Found OpenMP_CXX: -fopenmp (found version "4.5")
      -- Found OpenMP: TRUE (found version "4.5")
      -- x86 detected
      -- Adding CPU backend variant ggml-cpu: -march=native
      -- Found CUDAToolkit: /usr/local/cuda/include (found version "12.4.131")
      -- CUDA Toolkit found
      -- Using CUDA architectures: 50-virtual;61-virtual;70-virtual;75-virtual;80-virtual;86-real;89-real
      CMake Error at /usr/share/cmake-3.22/Modules/CMakeDetermineCompilerId.cmake:726 (message):
        Compiling the CUDA compiler identification source file
        "CMakeCUDACompilerId.cu" failed.
      
        Compiler: CMAKE_CUDA_COMPILER-NOTFOUND
      
        Build flags:
      
        Id flags: -v
        The output was:
      
        No such file or directory
      
      Call Stack (most recent call first):
        /usr/share/cmake-3.22/Modules/CMakeDetermineCompilerId.cmake:6 (CMAKE_DETERMINE_COMPILER_ID_BUILD)
        /usr/share/cmake-3.22/Modules/CMakeDetermineCompilerId.cmake:48 (__determine_compiler_id_test)
        /usr/share/cmake-3.22/Modules/CMakeDetermineCUDACompiler.cmake:298 (CMAKE_DETERMINE_COMPILER_ID)
        vendor/llama.cpp/ggml/src/ggml-cuda/CMakeLists.txt:43 (enable_language)
      
      
      -- Configuring incomplete, errors occurred!
      See also "/tmp/tmpwhq61gdi/build/CMakeFiles/CMakeOutput.log".
      See also "/tmp/tmpwhq61gdi/build/CMakeFiles/CMakeError.log".
      
      *** CMake configuration failed
      [end of output]
  
  note: This error originates from a subprocess, and is likely not a problem with pip.
  ERROR: Failed building wheel for llama_cpp_python
Failed to build llama_cpp_python
ERROR: Failed to build installable wheels for some pyproject.toml based projects (llama_cpp_python)

3. 确保 CUDA 环境正确配置

python 复制代码 
运行以下命令检查 CUDA 是否安装正确：
nvcc --version
如果报错 
nvcc: command not found
，说明 CUDA 环境未正确配置。需要：

安装 CUDA Toolkit（推荐 CUDA 12.x 或 11.x）

确保 nvcc 在 PATH 中：export PATH=/usr/local/cuda/bin:$PATH
python 复制代码 
$ nvcc -version
bash: nvcc: command not found
python 复制代码 
1. 确认是否安装了 CUDA Toolkit

运行以下命令检查 CUDA 是否安装：
ls /usr/local/cuda

如果目录不存在，说明 CUDA Toolkit 未安装，需要先安装它。
python 复制代码 
$ ls /usr/local/cuda
bin  compute-sanitizer  DOCS  EULA.txt  extras  gds  gds-12.4  include  lib64  libnvvp  nsight-compute-2024.1.1  nsightee_plugins  nsight-systems-2023.4.4  nvml  nvvm  README  share  src  targets  tools  version.json

目录结构来看，CUDA Toolkit 已经安装（版本可能是 12.4），但 nvcc 仍然不可用，可能是因为： nvcc 路径未添加到 PATH CUDA 环境变量未正确配置 安装过程中部分组件缺失

解决方案 1. 手动添加 nvcc 到 PATH

export PATH=/usr/local/cuda/bin:$PATH

然后验证： nvcc --version

python 复制代码 
将以下内容添加到 
~/.bashrc
 或 
~/.zshrc
：

export PATH=/usr/local/cuda/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH

然后生效：
source ~/.bashrc
python 复制代码 
$ nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2024 NVIDIA Corporation
Built on Thu_Mar_28_02:18:24_PDT_2024
Cuda compilation tools, release 12.4, V12.4.131
Build cuda_12.4.r12.4/compiler.34097967_0

4. 重新命令行安装 llama-cpp-python

确保 nvcc 可用后，

运行：

python 复制代码 
CMAKE_ARGS="-DGGML_CUDA=on -DCMAKE_CUDA_COMPILER=/usr/local/cuda/bin/nvcc" pip install llama-cpp-python

仍然报错

python 复制代码 
FAILED: vendor/llama.cpp/tools/mtmd/llama-llava-clip-quantize-cli
      : && /usr/bin/g++  -pthread -B /home/anaconda3/envs/llmtuner/compiler_compat -O3 -DNDEBUG  vendor/llama.cpp/tools/mtmd/CMakeFiles/llava.dir/llava.cpp.o vendor/llama.cpp/tools/mtmd/CMakeFiles/llava.dir/clip.cpp.o vendor/llama.cpp/tools/mtmd/CMakeFiles/llama-llava-clip-quantize-cli.dir/clip-quantize-cli.cpp.o -o vendor/llama.cpp/tools/mtmd/llama-llava-clip-quantize-cli  -Wl,-rpath,/tmp/tmptjbwss02/build/bin:  vendor/llama.cpp/common/libcommon.a  bin/libllama.so  bin/libggml.so  bin/libggml-cpu.so  bin/libggml-cuda.so  bin/libggml-base.so && :
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: warning: libgomp.so.1, needed by bin/libggml-cpu.so, not found (try using -rpath or -rpath-link)
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: warning: libcuda.so.1, needed by bin/libggml-cuda.so, not found (try using -rpath or -rpath-link)
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: warning: libdl.so.2, needed by /usr/local/cuda/lib64/libcudart.so.12, not found (try using -rpath or -rpath-link)
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: warning: libpthread.so.0, needed by /usr/local/cuda/lib64/libcudart.so.12, not found (try using -rpath or -rpath-link)
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: warning: librt.so.1, needed by /usr/local/cuda/lib64/libcudart.so.12, not found (try using -rpath or -rpath-link)
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemCreate'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cpu.so: undefined reference to `GOMP_barrier@GOMP_1.0'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemAddressReserve'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemUnmap'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cpu.so: undefined reference to `GOMP_parallel@GOMP_4.0'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemSetAccess'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuDeviceGet'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cpu.so: undefined reference to `omp_get_thread_num@OMP_1.0'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemAddressFree'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuGetErrorString'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cpu.so: undefined reference to `GOMP_single_start@GOMP_1.0'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuDeviceGetAttribute'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemMap'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemRelease'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cpu.so: undefined reference to `omp_get_num_threads@OMP_1.0'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemGetAllocationGranularity'
      collect2: error: ld returned 1 exit status
      [165/165] : && /usr/bin/g++  -pthread -B /home/anaconda3/envs/llmtuner/compiler_compat -O3 -DNDEBUG  vendor/llama.cpp/tools/mtmd/CMakeFiles/mtmd.dir/mtmd.cpp.o vendor/llama.cpp/tools/mtmd/CMakeFiles/mtmd.dir/clip.cpp.o vendor/llama.cpp/tools/mtmd/CMakeFiles/llama-mtmd-cli.dir/mtmd-cli.cpp.o -o vendor/llama.cpp/tools/mtmd/llama-mtmd-cli  -Wl,-rpath,/tmp/tmptjbwss02/build/bin:  vendor/llama.cpp/common/libcommon.a  bin/libllama.so  bin/libggml.so  bin/libggml-cpu.so  bin/libggml-cuda.so  bin/libggml-base.so && :
      FAILED: vendor/llama.cpp/tools/mtmd/llama-mtmd-cli
      : && /usr/bin/g++  -pthread -B /home/anaconda3/envs/llmtuner/compiler_compat -O3 -DNDEBUG  vendor/llama.cpp/tools/mtmd/CMakeFiles/mtmd.dir/mtmd.cpp.o vendor/llama.cpp/tools/mtmd/CMakeFiles/mtmd.dir/clip.cpp.o vendor/llama.cpp/tools/mtmd/CMakeFiles/llama-mtmd-cli.dir/mtmd-cli.cpp.o -o vendor/llama.cpp/tools/mtmd/llama-mtmd-cli  -Wl,-rpath,/tmp/tmptjbwss02/build/bin:  vendor/llama.cpp/common/libcommon.a  bin/libllama.so  bin/libggml.so  bin/libggml-cpu.so  bin/libggml-cuda.so  bin/libggml-base.so && :
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: warning: libgomp.so.1, needed by bin/libggml-cpu.so, not found (try using -rpath or -rpath-link)
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: warning: libcuda.so.1, needed by bin/libggml-cuda.so, not found (try using -rpath or -rpath-link)
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: warning: libdl.so.2, needed by /usr/local/cuda/lib64/libcudart.so.12, not found (try using -rpath or -rpath-link)
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: warning: libpthread.so.0, needed by /usr/local/cuda/lib64/libcudart.so.12, not found (try using -rpath or -rpath-link)
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: warning: librt.so.1, needed by /usr/local/cuda/lib64/libcudart.so.12, not found (try using -rpath or -rpath-link)
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemCreate'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cpu.so: undefined reference to `GOMP_barrier@GOMP_1.0'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemAddressReserve'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemUnmap'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cpu.so: undefined reference to `GOMP_parallel@GOMP_4.0'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemSetAccess'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuDeviceGet'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cpu.so: undefined reference to `omp_get_thread_num@OMP_1.0'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemAddressFree'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuGetErrorString'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cpu.so: undefined reference to `GOMP_single_start@GOMP_1.0'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuDeviceGetAttribute'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemMap'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemRelease'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cpu.so: undefined reference to `omp_get_num_threads@OMP_1.0'
      /home/anaconda3/envs/llmtuner/compiler_compat/ld: bin/libggml-cuda.so: undefined reference to `cuMemGetAllocationGranularity'
      collect2: error: ld returned 1 exit status
      ninja: build stopped: subcommand failed.
      
      *** CMake build failed
      [end of output]
  
  note: This error originates from a subprocess, and is likely not a problem with pip.
  ERROR: Failed building wheel for llama-cpp-python
Failed to build llama-cpp-python
ERROR: Failed to build installable wheels for some pyproject.toml based projects (llama-cpp-python)

从错误日志来看，主要问题是链接阶段找不到关键的 CUDA 和 OpenMP 动态库（ libcuda.so.1 、 libgomp.so.1 等），同时存在符号未定义的问题（如 cuMemCreate ）。以下是完整的解决方案：

5. 修复cuda so库

python 复制代码 
1. 修复系统库路径（关键步骤）

添加 CUDA 和系统库路径到环境变量

export LD_LIBRARY_PATH=/usr/local/cuda/lib64:/usr/lib/x86_64-linux-gnu:$LD_LIBRARY_PATH

永久生效：
echo 'export LD_LIBRARY_PATH=/usr/local/cuda/lib64:/usr/lib/x86_64-linux-gnu:$LD_LIBRARY_PATH' >> ~/.bashrc
source ~/.bashrc

验证库是否存在

ls /usr/lib/x86_64-linux-gnu/libgomp.so.1      # OpenMP 库
ls /usr/local/cuda/lib64/libcudart.so.12       # CUDA Runtime
ls /usr/lib/x86_64-linux-gnu/libcuda.so.1      # NVIDIA 驱动库
python 复制代码 
$ echo 'export LD_LIBRARY_PATH=/usr/local/cuda/lib64:/usr/lib/x86_64-linux-gnu:$LD_LIBRARY_PATH' >> ~/.bashrc
$ source ~/.bashrc                                                                                                                                                                                                                                                     
$ ls /usr/lib/x86_64-linux-gnu/libgomp.so.1 
/usr/lib/x86_64-linux-gnu/libgomp.so.1
$ ls /usr/local/cuda/lib64/libcudart.so.12
/usr/local/cuda/lib64/libcudart.so.12
$ ls /usr/lib/x86_64-linux-gnu/libcuda.so.1
/usr/lib/x86_64-linux-gnu/libcuda.so.1

6. 重新安装并指定完整 CMake 参数

python 复制代码 
CMAKE_ARGS="-DGGML_CUDA=ON \
            -DCMAKE_CUDA_COMPILER=/usr/local/cuda/bin/nvcc \
            -DCMAKE_CUDA_ARCHITECTURES=80 \
            -DLLAMA_CUDA_FORCE_DMMV=ON" \
  pip install llama-cpp-python \
    --force-reinstall \
    --no-cache-dir \
    --verbose \
    -i https://pypi.tuna.tsinghua.edu.cn/simple
python 复制代码 
最终可以看到：
Successfully installed MarkupSafe-3.0.2 diskcache-5.6.3 jinja2-3.1.6 llama-cpp-python-0.3.9 numpy-2.2.6 typing-extensions-4.14.0
相关推荐
明月_清风9 小时前
FastAPI 从入门到实战：3 分钟构建高性能异步 API
后端·python·fastapi
笨拙的老猴子9 小时前
[特殊字符] Java GC机制详解：G1、ZGC、Shenandoah全面解析与版本演进对比
java·开发语言
bellus-9 小时前
ubuntu26测试win10的ollama大模型性能
python
水木流年追梦9 小时前
大模型入门-Reward 奖励模型训练
开发语言·python·算法·leetcode·正则表达式
JavaWeb学起来9 小时前
Python学习教程（六）数据结构List(列表)
数据结构·python·python基础·python教程
liuyunshengsir9 小时前
PyTorch 动态量化（Dynamic Quantization）
人工智能·pytorch·python
电子云与长程纠缠9 小时前
UE5制作六边形包裹球体效果
开发语言·python·ue5
砍材农夫10 小时前
物联网 基于netty构建mqtt协议规范(遗嘱与保留消息)
java·开发语言·物联网·netty
DFT计算杂谈10 小时前
KPROJ编译教程
java·前端·python·算法·conda
froginwe1110 小时前
Python3 迭代器与生成器
开发语言
热门推荐
01GitHub 镜像站点02Codex 接入 DeepSeek API 完整配置文档03【踩坑记录 | 第一篇】微软商店无法使用时，如何手动安装 OpenAI Codex？附`.msix`文件系统错误解决方法04裂开！ChatGPT 居然开始要手机号验证，附详细解决方法05CC-Switch & Claude 基于 Linux 服务器安装使用指南06几个好用的ip纯净度检测网站07装上就回不去了：CodeGraph 让 AI 编程效率飙升 92%，它到底做了什么？08【AI】2026 年具身智能模型和世界模型总结09codex app每次打开重连5次Reconnecting问题解决10用了半年 OpenRouter，我换到了 Ofox.ai — 两个 AI API 聚合平台的真实对比